A hybrid neuromorphic processing architecture and event coordination method for 6g inter-signal

Abstract

The application is suitable for the technical field of 6G terminal signal processing, and provides a hybrid neuromorphic processing architecture and event coordination method for 6G sensing signals, the architecture comprising: a dedicated pulse encoder, used for operating a time-frequency spectrum or a range-Doppler spectrum converted from a sensing baseband signal, and triggering a pulse event according to a dynamic threshold θ=α·N+β·I; a lightweight pulse neural network, used for performing primary pattern recognition on a pulse event stream; a programmable intelligent arbitrator, connected to the lightweight pulse neural network, used for making a decision on a primary recognition result according to a preset or configurable rule; a raw signal circular buffer, used for continuously storing raw high-sampling-rate I / Q data in a recent period of time; and a digital processing channel, in a default dormant state, used for performing high-precision signal analysis. The application solves the failure problem of a general neuromorphic scheme in a low-SNR multipath scene, and realizes high-reliability ultra-low-power sensing.

Description

Technical Field

[0001] This invention relates to the field of 6G terminal signal processing technology, specifically to a hybrid neuromorphic processing architecture and event coordination method for 6G sensory signals. Background Technology

[0002] The application of neuromorphic computing in low-power sensing is a known trend. However, applying it to 6G sensing signal processing is not a simple application migration. Sensing signals have unique sparse structures and dynamic ranges in the range-Doppler or time-frequency domains, and their operating environments are often accompanied by low signal-to-noise ratios (SNR) and multipath interference. Existing general-purpose neuromorphic solutions (such as pulse coding based on visual events) directly processing such signals can lead to serious problems: insensitivity to weak targets resulting in missed pulses, or excessive sensitivity to noise and multipath interference leading to pulse flooding. Existing hybrid architectures lack specific consideration for the characteristics of the sensing physical layer; their event generation rules do not match the signal characteristics, failing to guarantee low false alarms and low false wake-ups in real wireless environments, thus negating energy-saving benefits. There are two types of coupling failure modes in sensing scenarios: 1. Low SNR leads to insufficient energy changes in weak targets, causing missed pulses; 2. Multipath interference and broadband interference cause abnormal local energy fluctuations, leading to pulse flooding and false wake-ups. General-purpose solutions struggle to suppress both types of failures simultaneously. The rules for generating synesthetic events and the arbitration wake-up logic need to be designed collaboratively: if event generation is not customized for the distance-Doppler / time-frequency structure, the arbitration rules cannot set stable physical semantic thresholds; if arbitration relies solely on SNN output without a raw data verification path with timestamp windowing, accuracy and energy efficiency cannot be simultaneously satisfied during multipath dynamic changes. Therefore, a hybrid neuromorphic processing architecture and event coordination method for 6G synesthetic signals are needed to address these issues. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a hybrid neuromorphic processing architecture and event coordination method for 6G sensing signals, thereby resolving the problems mentioned in the background. This architecture does not simply combine existing neuromorphic and digital modules; instead, it forms a deeply integrated closed loop through innovative dedicated pulse coding rules, matching arbitration wake-up logic, and data extraction mechanisms, ensuring excellent energy efficiency and accuracy even in harsh wireless environments.

[0004] This invention is implemented as follows: a hybrid neuromorphic processing architecture for 6G sensory signals, the architecture comprising: A dedicated pulse encoder is used to operate on the time spectrum or range-Doppler spectrum of the converted syn-inductive baseband signal. It triggers pulse events based on a dynamic threshold θ=α·N+β·I, where N is the local noise basis estimate, I is the current channel broadband interference level, and α and β are weighting coefficients. A lightweight pulse neural network, connected to the dedicated pulse encoder, is used for primary pattern recognition of pulse event streams; A programmable intelligent arbitrator, connected to the lightweight spiking neural network, is used to adjudicate the initial identification results according to preset or configurable rules; Raw signal cyclic buffer is used to continuously store raw high-sampling-rate I / Q data from the most recent period; The digital processing path is in a dormant state by default and is used to perform high-precision signal analysis. When the programmable intelligent arbitrator determines an event to be valid, it extracts the original I / Q data corresponding to the time window from the original signal cyclic buffer based on the event timestamp, wakes up the digital processing path for analysis, and then immediately puts the digital processing path into sleep mode.

[0005] As a further aspect of the present invention, the dynamic threshold θ is constructed to simultaneously suppress leakage pulses under low signal-to-noise ratio conditions and pulse flooding under multipath interference conditions.

[0006] As a further aspect of the present invention, the arbitration rules of the programmable intelligent arbitrator can be updated via a software interface or configured by the network side.

[0007] As a further aspect of the present invention, the arbitration rules of the programmable intelligent arbitrator include judging the temporal continuity, spatial distribution, or specific pattern of pulse events.

[0008] As a further aspect of the present invention, the architecture maintains only low-power operation of the dedicated pulse encoder, the lightweight spiking neural network, and the original signal cyclic buffer during invalid events, while the digital processing path is in a deep sleep state.

[0009] As a further aspect of the present invention, when the dedicated pulse encoder analyzes the range-Doppler spectrum, it can adaptively adjust the threshold according to the broadband interference level I to filter out static clutter interference.

[0010] Another objective of this invention is to provide a hybrid neuromorphic event coordination method for 6G synesthetic signals, the method comprising the following steps: Transform the syn-inductive baseband signal to obtain the time spectrum or range-Doppler spectrum; The spectrum is pulse-coded using a dedicated pulse encoder. The pulse triggering condition is a dynamic threshold θ = α·N + β·I, where N is the noise floor estimate, I is the broadband interference level, and α and β are weighting coefficients. The pulse event is input into a lightweight spiking neural network for initial recognition, and the preliminary recognition result is output. The programmable intelligent arbitrator adjudicates the initial identification results to determine whether they are valid events. If the event is determined to be valid, the original I / Q data of the corresponding time window is extracted from the original signal cyclic buffer according to the event timestamp, and the digital processing path is activated for high-precision analysis. After the high-precision analysis is completed, the digital processing path automatically returns to sleep mode.

[0011] As a further aspect of the present invention, the adjustment of the dynamic threshold θ is used to lower the threshold to capture weak targets when the signal-to-noise ratio is low, and to raise the threshold to suppress false alarms when there is multipath interference.

[0012] As a further aspect of the present invention, the arbitration rules of the programmable intelligent arbitrator are preset for different application scenarios and can be remotely configured.

[0013] As a further aspect of the present invention, the method analyzes the range-Doppler spectrum in a vehicle blind spot collision avoidance monitoring scenario and determines pedestrian targets through continuous pulse sequence pattern recognition.

[0014] Compared with the prior art, the beneficial effects of the present invention are: A dedicated pulse encoder designed for the physical layer characteristics of syn-sensory signals triggers pulse events based on a dynamic threshold θ=α·N+β·I. It accurately extracts raw data through time-spectrum / distance-Doppler spectrum operations, environmental adaptive thresholds, and timestamp-driven methods, solving the failure problem of general neuromorphic solutions in low SNR multipath scenarios. It achieves highly reliable ultra-low power sensing and can simultaneously suppress two types of coupling failures: low SNR leaked pulses and multipath pulse flooding. Attached Figure Description

[0015] Figure 1 This is a structural diagram of a hybrid neuromorphic processing architecture for 6G sensory signals.

[0016] Figure 2 This is a flowchart of the workflow of a dedicated pulse encoder in a hybrid neuromorphic processing architecture for 6G sensor signals.

[0017] Figure 3 This is a working closed-loop diagram of a hybrid neuromorphic processing architecture for 6G sensor signals. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0020] like Figure 1 , Figure 2 and Figure 3 As shown, this embodiment of the invention provides a hybrid neuromorphic processing architecture for 6G sensory signals, the architecture comprising: A dedicated pulse encoder is used to operate on the time spectrum or range-Doppler spectrum of the converted baseband signal. It triggers pulse events based on a dynamic threshold θ = α·N + β·I. The trigger occurs when the change in unit energy exceeds the dynamic threshold θ, where N is the estimated local noise basis, I is the current channel broadband interference level, and α and β are weighting coefficients. This mechanism ensures that the encoder can adapt to the environment, lowering the threshold to capture weak targets at low SNR (signal-to-noise ratio) and raising the threshold to suppress false alarms in the event of multipath interference. A lightweight spiking neural network (SNN) is connected to the dedicated pulse encoder for primary pattern recognition of the pulse event stream; A programmable intelligent arbitrator, a state machine with updatable rules, adjudicates the output of the SNN and is connected to the lightweight spiking neural network to adjudicate the primary recognition results according to preset or configurable rules; Raw signal cyclic buffer is used to continuously store raw high-sampling-rate I / Q data from the most recent period; The digital processing path is in a dormant state by default and is used to perform high-precision signal analysis.

[0021] In this embodiment of the invention, when the programmable intelligent arbitrator determines an event to be valid, it extracts the original I / Q data corresponding to the time window from the original signal cyclic buffer based on the event timestamp, wakes up the digital processing path for analysis, and then immediately puts the digital processing path into sleep mode. If a general amplitude-pulse coding mechanism is used instead of the N / I joint dynamic threshold mechanism of this embodiment, pulse flooding or loss of weak targets will occur in low SNR multipath scenarios, leading to a sharp increase in the false wake-up rate of the arbitrator, thus completely negating the energy-saving benefits brought by neuromorphic computing.

[0022] In this embodiment of the invention, the dynamic threshold θ is constructed to simultaneously suppress leaked pulses under low signal-to-noise ratio conditions and pulse flooding under multipath interference conditions.

[0023] In this embodiment of the invention, the arbitration rules of the programmable intelligent arbitrator can be updated via a software interface or configured by the network side. The programmable arbitrator enables the system to adapt to different application scenarios, from consumer electronics to the Industrial Internet of Things, through software updates, thus protecting hardware investments. The arbitration rules of the programmable intelligent arbitrator include judgments on the temporal continuity, spatial distribution, or specific patterns of pulse events.

[0024] In this embodiment of the invention, the architecture maintains low-power operation of only the dedicated pulse encoder, lightweight spiking neural network, and raw signal cyclic buffer during invalid events, while the digital processing path is in a deep sleep state. When the dedicated pulse encoder analyzes the range-Doppler spectrum, it can adaptively adjust the threshold according to the broadband interference level I to filter out static clutter interference.

[0025] The embodiments of this invention realize a common design closed loop for the physical layer: instead of modular assembly, it is a full-chain customization from event generation (θ=α·N+β·I), SNN initial judgment, programmable decision, to timestamp windowing and precise wake-up, ensuring the system's unique optimization for synesthesia signals, while suppressing two types of failures: low SNR leaked pulses and multipath pulse flooding.

[0026] The embodiments of this invention demonstrate practical reliability and a robust trend in energy efficiency improvement: In an indoor channel environment with 5GHz bandwidth, dynamic SNR variations from -10dB to 10dB, and typical multipath propagation, compared to a continuously running traditional DSP solution, this architecture achieves an average power consumption reduction of over 98% (i.e., two orders of magnitude) while maintaining a target detection rate greater than 95%. Its energy efficiency advantage remains robust even in complex wireless environments. The above results are exemplary prototype test data, with test conditions including: 5GHz signal bandwidth, dynamic SNR environment, and typical indoor multipath channel; the baseline is a continuously running traditional digital signal processor solution. Specific values ​​may vary depending on hardware technology, scenario, and signal parameters. The energy efficiency and detection rate advantages maintain a consistent trend in complex wireless environments with simultaneous dynamic SNR and multipath propagation, and the advantages of the embodiments of this invention become more significant as target density and background interference increase.

[0027] The embodiments of the present invention can also effectively defend against application migration: by emphasizing the combination of the special technical features of "time spectrum / distance-Doppler spectrum operation" and "dynamic threshold θ=α·N+β·I", a clear technical boundary is formed with general neuromorphic schemes.

[0028] This invention also provides a hybrid neuromorphic event coordination method for 6G synergistic signals, the method comprising the following steps: Step 1: Transform the syn-inductive baseband signal to obtain the time spectrum or range-Doppler spectrum; Step 2: Use a dedicated pulse encoder to pulse encode the spectrum. The pulse triggering condition is a dynamic threshold θ = α·N + β·I, where N is the noise floor estimate and I is the broadband interference level. Step 3: Input the pulse event into a lightweight spiking neural network for initial recognition and output preliminary recognition results; Step 4: The programmable intelligent arbitrator adjudicates the initial identification results to determine whether they are valid events; Step 5: If the event is determined to be valid, the original I / Q data corresponding to the time window is extracted from the original signal cyclic buffer according to the event timestamp, and the digital processing path is activated for high-precision analysis. Step 6: After the high-precision analysis is completed, the digital processing path automatically returns to sleep mode, waiting for the next wake-up.

[0029] In this embodiment of the invention, the adjustment of the dynamic threshold θ is used to lower the threshold to capture weak targets at low signal-to-noise ratios and to raise the threshold to suppress false alarms during multipath interference. The adjudication rules of the programmable intelligent arbitrator are preset for different application scenarios (such as indoor personnel monitoring and vehicle collision avoidance) and can be remotely configured.

[0030] This invention is applied to vehicle blind spot collision avoidance monitoring, where the terminal continuously processes 77GHz vehicle radar sensing signals. A dedicated pulse encoder analyzes the range-Doppler spectrum, effectively filtering out road surface clutter (higher I, higher θ) interference using an N / I dynamic threshold. When a pedestrian enters the blind spot, their specific movement pattern generates a regular pulse sequence on the spectrum, and the pedestrian target is identified through continuous pulse sequence pattern recognition. The arbitrator (with the preset rule of "three consecutive pulses matching the pedestrian's Doppler characteristics") determines it as a valid event, immediately extracts the raw data from the buffer for the preceding and following 0.5 seconds, and activates the digital pathway for high-resolution imaging and hazard level determination. 99% of the time, only the neuromorphic pathway operates at milliwatt-level power consumption, and the digital pathway is only activated momentarily in critical moments, achieving an ultimate balance between safety and energy efficiency.

[0031] In a real-world test environment, based on a typical indoor channel setting (5GHz bandwidth, dynamic SNR), this solution exhibits the following performance metrics: 1. Improved energy efficiency: Compared with traditional continuously operating DSP solutions, the average power consumption is reduced by more than 98%.

[0032] 2. Detection rate maintenance: In complex wireless environments, the target detection rate remains greater than 95%, which is significantly higher than that of general neuromorphic solutions (approximately 65%).

[0033] 3. Response latency: The end-to-end latency from valid event determination to completion of analysis in the digital pathway is less than 50ms.

[0034] 4. Resource consumption: The dedicated pulse encoder and lightweight SNN have a memory consumption of less than 50MB, making them suitable for embedded platforms.

[0035] The above description only details the preferred embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0036] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0037] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A hybrid neuromorphic processing architecture for 6G sensory signals, characterized in that, The architecture includes: A dedicated pulse encoder is used to operate on the time spectrum or range-Doppler spectrum of the converted syn-inductive baseband signal. It triggers pulse events based on a dynamic threshold θ=α·N+β·I, where N is the local noise basis estimate, I is the current channel broadband interference level, and α and β are weighting coefficients. A lightweight pulse neural network, connected to the dedicated pulse encoder, is used for primary pattern recognition of pulse event streams; A programmable intelligent arbitrator, connected to the lightweight spiking neural network, is used to adjudicate the initial identification results according to preset or configurable rules; Raw signal cyclic buffer is used to continuously store raw high-sampling-rate I / Q data from the most recent period; The digital processing path is in a dormant state by default and is used to perform high-precision signal analysis. When the programmable intelligent arbitrator determines an event to be valid, it extracts the original I / Q data corresponding to the time window from the original signal cyclic buffer based on the event timestamp, wakes up the digital processing path for analysis, and then immediately puts the digital processing path into sleep mode.

2. The hybrid neuromorphic processing architecture for 6G sensory signals according to claim 1, characterized in that, The dynamic threshold θ is constructed to simultaneously suppress leaked pulses under low signal-to-noise ratio conditions and pulse flooding under multipath interference conditions.

3. The hybrid neuromorphic processing architecture for 6G sensory signals according to claim 1, characterized in that, The arbitration rules of the programmable intelligent arbitrator can be updated via a software interface or configured by the network side.

4. The hybrid neuromorphic processing architecture for 6G sensory signals according to claim 1, characterized in that, The arbitration rules of the programmable intelligent arbitrator include judging the temporal continuity, spatial distribution, or specific patterns of pulse events.

5. The hybrid neuromorphic processing architecture for 6G sensory signals according to claim 1, characterized in that, During invalid events, the architecture maintains low-power operation of only the dedicated pulse encoder, lightweight spiking neural network, and raw signal cyclic buffer, while the digital processing path is in a deep sleep state.

6. The hybrid neuromorphic processing architecture for 6G sensory signals according to claim 1, characterized in that, When the dedicated pulse encoder analyzes the range-Doppler spectrum, it can adaptively adjust the threshold according to the broadband interference level I to filter out static clutter interference.

7. A hybrid neuromorphic event coordination method for 6G synergistic signals, characterized in that, The method includes the following steps: Transform the syn-inductive baseband signal to obtain the time spectrum or range-Doppler spectrum; The spectrum is pulse-coded using a dedicated pulse encoder. The pulse triggering condition is a dynamic threshold θ = α·N + β·I, where N is the noise floor estimate, I is the broadband interference level, and α and β are weighting coefficients. The pulse event is input into a lightweight spiking neural network for initial recognition, and the preliminary recognition result is output. The programmable intelligent arbitrator adjudicates the initial identification results to determine whether they are valid events. If the event is determined to be valid, the original I / Q data of the corresponding time window is extracted from the original signal cyclic buffer according to the event timestamp, and the digital processing path is activated for high-precision analysis. After the high-precision analysis is completed, the digital processing path automatically returns to sleep mode.

8. The hybrid neuromorphic event coordination method for 6G synergistic signals according to claim 7, characterized in that, The adjustment of the dynamic threshold θ is used to lower the threshold to capture weak targets at low signal-to-noise ratios and to raise the threshold to suppress false alarms during multipath interference.

9. The hybrid neuromorphic event coordination method for 6G synergistic signals according to claim 7, characterized in that, The programmable intelligent arbitrator's arbitration rules are preset for different application scenarios and can be configured remotely.

10. The hybrid neuromorphic event coordination method for 6G synergistic signals according to claim 7, characterized in that, The method analyzes the range-Doppler spectrum in the vehicle blind spot collision avoidance monitoring scenario and identifies pedestrian targets through continuous pulse sequence pattern recognition.

Images